Project Summary Suicide is one of the greatest challenges facing Society today. Although research to date has provided important insight into the risk factors for suicide, it has struggled to make a translational impact on preventing suicide. New research strategies are therefore vital. The RDoC framework offers a promising approach to researching suicide, by moving the focus away from mental illness as the primary predictor of suicide, towards an emphasis on transdiagnostic dimensions, novel predictors, and integration of multiple units of analysis. Identifying the neurobiology underlying the transdiagnostic risk factors for suicide is an important step in understanding the mechanisms that lead to suicide. The majority of neuroimaging research to date has been purely correlational in nature, and often fails to generalize outside of the study sample. Predictive modelling, on the other hand, uses cross-validation to increase generalizability, and allows for the prediction of behavior, which is likely to have greater real-world utility. In this proposal, we will use the RDoC framework, multimodal imaging and predictive modelling to identify the molecular and network-level underpinnings of risk factors for suicide in a large, transdiagnostic dataset. Specifically, we aim to use Connectome-based Predictive Modelling (CPM) to generate and validate predictive models of suicide risk factors (depression severity, impulsivity, executive dysfunction and sleep disturbances), transdiagnostically, based on existing functional connectivity data (aim 1). Further, convergent research implicates glutamate dysfunction in suicidality, and emerging evidence suggests a specific role for the metabotropic glutamate receptor 5 (mGluR5) in suicide and suicide-related behaviors. Subjects in the existing dataset also participated in PET imaging with and [18F]FPEB, a radioligand specific for the metabotropic glutamate receptor 5 (mGluR5). We will investigate the association between mGluR5 and risk factors for suicide transdiagnostically (aim 2). Finally, we propose to integrate fMRI and PET data by using PET-weighted CPM to predict risk factors for suicide transdiagnostically (aim 3). Here, we will assess whether combining molecular and circuit-level units of analysis increases predictive performance and identifies novel predictive networks. In summary, we propose to use cutting-edge multimodal imaging, computational modelling and the RDoC framework in an innovative approach to gain a more comprehensive understanding of the neural mechanisms underlying risk factors for suicide across psychiatric disorders. Findings of the proposed study have the potential to inform risk models and identify novel molecular and network-level targets for preventative and treatment strategies that are critically needed to reduce the suicide rate.